Counter-Strike Map-Making in Source 2: A Hypothetical Concept

[Shima33]

So, a few hours ago, I was binge-watching video footage of The Lab demo's that came with the HTC Vive, and one of the demos that really impressed me was Vesper Peak - not because of it being anything spectacular from a VR point of view, but because of a technology used to acheive it's photo-realistic look, known as photogrammetry. The basic idea of photogrammetry is that you take a large amount of photo's of a certain object or enviroment and then run it through a specialist peice of software, which stitches all of the images together and creates a polygon/voxel hybrid model, which can then be directly texture mapped from the images. 

 

After seeing it in use in the Vesper Peak demo, I looked around for other games that used this photogrammetry technique, and found a little gem from 2014 known as The Vanishing Of Ethan Carter, an example image of which is in the spoiler:

Spoiler

De_Sparity's Source 2 port? Perhaps.

As you can see, the realism is truly startling, to the point where it actually feels genuine. If you're not convinced, find gameplay footage of The Vanishing of Ethan Carter, or Vesper Peak, and compare it today's more common graphical offerings - a real bump up in graphics, as far as I'm concerned.

 

Now, here's where my mind ran wild with the idea. I'm not sure if Vesper Peak runs on Unity or Source 2, but for the sake of argument, let's say Vesper Peak is a Source 2 tech demo. Now, let's say that a Counter-Strike game got made for Source 2. Not nessescarily a Source 2 port of CS:GO, but simply a Counter-Strike game, similar in style to CS:GO. Imagine photogrammatic locations in CS:GO, and how beautiful they would look.

 

That followed on into another idea - how would you create photogrammatic locations for this Source 2 version of CS, anyway? You could use real-life locations, sure, but then the layouts and overall structure of the map would be compromised for artistic beauty, which would greatly degenerate a game like CS. So, after giving it some thought, I came up with a middle-ground solution - physically making a ratio model of the map by hand, similar to how Id Software did for some of the sprites in DOOM, and then using photogrammetry on that model instead.

 

Design of a map would begin as usual - loading up whatever the Source 2 version of Hammer is, brainstorming a layout and aesthetic, etc., and playtesting it to work out all the kinks and bugs. Then, once you have a layout pinned down and a rough design/idea of the map's aesthetic, you would make a 1/16, 1/32 or 1/64-sized model mock-up of what your finished level should look like, and then snap hundreds of pictures of the physical model. The model would be made in sections, so that if there is a part of the level that is underground or in an inside area, you can simply detach the roofing so as to get a photogrammatic look of the insides. The result would be a map that directly emulates the layout and aesthetics of it's digital counterpart, but has the aesthetic beauty of a physically-mapped photogrammatic location. 

Further edits in order to fix gameplay and bugs can then be made in-engine to the photogrammatic model. Another thing that came to mind at the time of writing this is how much detail some fantasy roleplayers put into their scenery props and character models in tabletop games such as Warhammer - the physical mock-up model would look similar to that, in a sense.

 

As always, feedback and opinions appreciated. I'm really excited about this photogrammetry technology, and I hope to see more games in the near future that utilize it well.

 

Thanks for reading,

Shima.

Edited April 11, 2016 by Shima33
Typo.

[sevin]

That photogrammetry thing is neat. I've had that Ethan Carter game on my wishlist for awhile, never knew about photogrammetry. 

Is there any consumer software out now that can do this photogrammetry thing? 

As for your hypothesis, I don't think we'll be able to test it for many more years. Source 2 news has all but stopped for anything but Dota (correct?), so I'd guess we aren't even getting it for another 5 years. Maybe they'll introduce it with a new CS, like they did for Source with HL2. Who knows. 

This would be super cool though, I want to get Ethan Carter even more now.

[leplubodeslapin]

Ok first of all : play the vanishing of Ethan Carter. It's just great.

@Sjonsson posted a very nice video explaining what photogrametry is and showing the process to make 1 mesh using this technique :

The thing is, you don't have to fully respect the original environment to make your map using photogrametry technique. I'm pretty sure the environment of The Vanishing of Ethan Carter is not exactly the same as the real world reference (in terms of layout). You can use this technique for objects and textures, and then mix classic modeling methods with some objects being made with photogrametry.

I guess we will see more of this in future games.

[blackdog]

Slightly related, the other day a spammer posted about their 3D software: it works by mixing laser scans to pictures and the results seen in the video were interesting.

As for this technique, like leplubo I think you use it as a base, get a head start and then I'm sure you have to clean up models for a more efficient performance and position around the environment in a way that serves the gameplay. 

Also at some E3 or GDC in the last two years there was an FPS that looked ultra-realistic, was called project-something I believe. I think to remember that it had to do with photogrammetry.

[biXen]

Epic have some pretty good walkthroughs of how they made their assets for one of their presentation levels (mountainous balloon thing, can't remember which version). Check out their Youtube. There's also tutorials at for instance fxphd that are very in-depth.

[DrywallDreams]

7 hours ago, blackdog said:

Also at some E3 or GDC in the last two years there was an FPS that looked ultra-realistic, was called project-something I believe. I think to remember that it had to do with photogrammetry.

There's a thread for it somewhere around here, the level design was a bunch of empty rooms, from what I could see.

That's sort of the issue with photogrammetry: it turns out that the real world's kinda boring.  Photogrammetry is great if you want a ton of rocks and statues, but I'm not convinced that it's the future of content creation.

[blackdog]

So... would have worked great for Rage?

[blackdog]

Haven't seen this yet, but GDC just published this

 

[The Horse Strangler]

Photogrammetry is nice, but it's a very long and complicated process for a small team of artists to pull off, more specifically a team of 1/2 artists working on a csgo map. Photogrammetry reaps really nice benefits when working on larger scale projects where tons of art is needed (i.e save time on sculpting/highpoly because you don't need to do that anymore), but otherwise the process requires a lot of commitment and thus is generally not worth it at all.

Not only do you need to have some decent camera gear, but physically capturing locations can prove very challenging both budget and time wise. Lighting generally needs to be at a very ideal range, typically overcast or low ambient. Committing to photogrammetry also creates the problem of... committing to photogrammetry. When you start setting the bar at that level of res and surface definition you can't just go and construct a bunch of stuff the old way. The contrast between the definition of assets becomes incredibly jarring. As an individual it's a lot of time consuming work to take GOOD photos ( you can't just snap a bunch of random photos in a 360 degree rotation and input that into the software without worry), then to move on and retopo/ reuv etc.

Furthermore on a topic about Source 2, I honestly don't see csgo getting ported anytime soon. Realistically our current workflows wouldn't benefit much from other than the minor speedups modernizing some of the tools has provided. You're still dealing with vmt's (vmats), you will still need to compile lighting unless csgo moves to deferred rendering (unlikely), and for the most part producing art is more or less the same. Qc's are gone, and that's cool, but source 2 just isn't the day to night contrast that it really needs to be atm.

Edited August 18, 2016 by MrTwoVideoCards

[blackdog]

I wanted to post but need to do some search, on a new photogrammetry app for drones. Came up one late night I was youtubing (yes you get the most original ads sometimes) and was a full 10-15 min video demoing a drone flight and subsequent assembly of the imagery, showing different flight paths give different results.

I dunno, to me -a hobbyist- looking at this stuff makes me think that even for big teams this is a "waste of time". I mean, scanning areas take hours, and I can't imagine how long it takes to optimise all?

[The Horse Strangler]

1 hour ago, blackdog said:

I wanted to post but need to do some search, on a new photogrammetry app for drones. Came up one late night I was youtubing (yes you get the most original ads sometimes) and was a full 10-15 min video demoing a drone flight and subsequent assembly of the imagery, showing different flight paths give different results.

I dunno, to me -a hobbyist- looking at this stuff makes me think that even for big teams this is a "waste of time". I mean, scanning areas take hours, and I can't imagine how long it takes to optimise all?

It saves time for larger teams because generally you have a larger party of photographers/artists go out on location so you can make the most out of the trip. The greatest thing about photogrammetry of course is the natural ability to provide automatic lighting data (normal maps, gloss, etc) but because you get a lot of those micro details you generally skip the highpoly/sculpt phase on a lot of stuff. Looking at star wars battlefront, Endor is a really solid example because of how much variation and color you're able to pick out from the photos taken, stuff like this would have not only required a lot of sculpt time, but a lot of additional masks would have been authored to create tinting variation of moss and blending.

I would generally say that not EVERYTHING should be sourced from photogrammetry, but it really helps in areas where natural weathering has occurred and layers upon layers of variation in color gradation and buildup as well. That's a ton of stuff you get for free that would suck up a ton of hours if you needed to do that on each individual asset by hand.

[biXen]

To be honest I think it makes PERFECT sense for some situations whether you are one person or 50. But it makes more sense for movies because they already do all this work on a location already, most movies nowadays will Lidar scan, pick HDRis and shoot a lot of footage from a location. So running around and shooting assets makes sense too.

For a game set I only think it makes sense if a location has a lot of the assets you are after. As you aren't really confined to recreating the location like you are on a film set. For Unreal's demo it made sense to spend a few days doing all these "highlands" assets. They definitely saved time and upped the quality. There are some good workflows for it, but I think it would take too much time as a hobbyist if you are picking one object here and there and have to remember the process every time.

[blackdog]

With the spreading of photography tools is this becoming more accessible also for smaller teams? Like I'm thinking, a 360° camera can help? Then all those GoPro arrays…

[MaikelOrtega]

7 minutes ago, blackdog said:

With the spreading of photography tools is this becoming more accessible also for smaller teams? Like I'm thinking, a 360° camera can help? Then all those GoPro arrays…

I might be wrong, but I think the stitching needed in 360º images would do more harm that good in this case. And I don't think this could become a standard for smaller teams, it seems way too expensive. Requires lots of hours of fieldwork, in really specific conditions (you depend a lot on the weather, time of day, etc... As light should be stable). The results are astounding though. 

The guys working on The Vanishing of Ethan Carter actually made a post about this, is pretty interesting:
http://www.theastronauts.com/2014/03/visual-revolution-vanishing-ethan-carter/
 

[blackdog]

Will give it a read ?

Yeah for the conditions I found out first hand about overcast being best condition, when I was doing my own textures for HL1